Post on 19-Oct-2020
ORIGINALARTICLE
The geography of speciation in narrow-range endemics of the ‘Haenydra’ lineage(Coleoptera, Hydraenidae, Hydraena)
Ignacio Ribera1,2*, Agustın Castro3, Juan A. Dıaz4, Josefina Garrido5,
Ana Izquierdo2, Manfred A. Jach6 and Luis F. Valladares7
1Institute of Evolutionary Biology (CSIC-UPF),
Passeig Maritim de la Barceloneta 37, 08003
Barcelona, Spain, 2Departamento de
Biodiversidad y Biologıa Evolutiva, Museo
Nacional de Ciencias Naturales (CSIC),
Madrid, Spain, 3IES Clara Campoamor,
Lucena, Cordoba, Spain, 4Departamento de
Biologıa Animal, Universidad de Santiago,
Lugo, Spain, 5Departamento de Ecologıa y
Biologıa Animal, Universidad de Vigo, Vigo,
Spain, 6Naturhistorisches Museum Wien,
Wien, Austria, 7Departamento de
Biodiversidad y Gestion Ambiental (Zoologıa),
Universidad de Leon, Leon, Spain
*Correspondence: Ignacio Ribera, Institute of
Evolutionary Biology (CSIC-UPF), Passeig
Maritim de la Barceloneta 37, 08003 Barcelona,
Spain.
E-mail: ignacio.ribera@ibe.upf-csic.es
ABSTRACT
Aim We test whether species of western Mediterranean aquatic Coleoptera of the
‘Haenydra’ lineage (Hydraenidae, Hydraena) originated through: (1) successive
periods of dispersal and speciation, (2) range fragmentation by random
vicariance, or (3) range fragmentation by geographic isolation owing to a
general reduction of population density.
Location Europe.
Methods To discriminate between scenarios we use contrasting predictions of
the relationship between phylogenetic and geographic distance. The phylogeny
was based on 3 kb of four mitochondrial and two nuclear gene fragments of
about half of the known species of ‘Haenydra’, including most western
Mediterranean taxa. Divergences were estimated using a molecular clock. The
relationship between phylogenetic and geographic distance was tested using
bivariate plots, Mantel tests and comparison of the observed phylogeny with the
one minimizing geographic distances between species, as measured using
Euclidean minimum spanning trees (EMSTs).
Results The monophyly of ‘Haenydra’ was strongly supported, although its
phylogenetic placement was not resolved. ‘Haenydra’ was estimated to be of late
Miocene age, with most species originating during the Pleistocene. In two clades
(Hydraena tatii and Hydraena emarginata clades) there was a significant
association between geographic and phylogenetic distance, and the
reconstructed phylogeny was identical to that obtained through the EMST,
demonstrating a strong non-randomness of the geographic distribution of the
species. In two other clades (Hydraena iberica and Hydraena bitruncata clades)
there was no association between geographic and phylogenetic distance, and the
observed phylogeny was not the one minimizing geographic distances. In one of
the clades this seems to be due to a secondary, recent range expansion of one
species (H. iberica), which erased the geographic signal of their distributions.
Main conclusions We show that it is possible to obtain strong evidence of stasis
of the geographic ranges of narrow-range endemic species through the study of
their phylogenetic relationships and current distributions. In at least two of the
studied clades, current species seem to have originated through the fragmentation
of a more widely distributed species, without further range movements. A process
of range expansion and fragmentation may have occurred repeatedly within the
‘Haenydra’ lineage, contributing to the accumulation of narrow-range endemics
in Mediterranean Pleistocene refugia.
Keywords
Aquatic Coleoptera, Hydraenidae, Iberian Peninsula, narrow-range endemics,
Pleistocene refugia, range expansion, speciation.
Journal of Biogeography (J. Biogeogr.) (2011) 38, 502–516
502 http://wileyonlinelibrary.com/journal/jbi ª 2010 Blackwell Publishing Ltddoi:10.1111/j.1365-2699.2010.02417.x
INTRODUCTION
Since the early definitions of speciation modes as sympatric,
allopatric or peripatric (Mayr, 1963) the fundamental role of
geography has been recognized, and there have been many
attempts to reconstruct the history of speciation through the
distributions of current species (e.g. Lynch, 1989; Barraclough
& Vogler, 2000; Fitzpatrick & Turelli, 2006). There is, however,
a recognized weakness common to all these studies: species do
change their geographic ranges, and it cannot be assumed that
the current geographic range of a species is the same as that at
the time of speciation, or that ranges are maintained through
the cladogenetic process (Gaston, 2003). This has prompted
many authors to conclude that evolutionary inferences con-
cerning the geography of species in the past will often not be
reliable (Chesser & Zink, 1994; Gaston, 1998; Losos & Glor,
2003). It would be equally wrong, however, to assume that all
species have suffered modifications in their ranges large
enough to erase any geographic signal from the past, as in
some cases there is strong evidence to support the stasis of
geographic ranges, either through the fossil record (e.g.
Jablonski, 1987) or with indirect evidence from ecological
and phylogenetic data (e.g. Carranza & Wade, 2004; Martınez-
Solano et al., 2004). Lineages with an abundance of narrowly
distributed, mostly allopatric species are particularly difficult
cases. The reduced range (often the result of specialized
ecological requirements) and non-overlapping distribution,
sometimes through several cladogenetic events (Fitzpatrick &
Turelli, 2006), strongly suggest allopatric speciation, but one
then has to ask whether the species originated, and have always
persisted, where they are currently found.
A possible way to test the persistence of a geographic signal
in the current distribution of a clade of species is through the
comparison of observed phylogenetic and spatial relationships
with a random null model (Barraclough & Nee, 2001). Using
this approach, we test here three potential scenarios for the
origin of several clades with mostly allopatric, narrowly
distributed species in a genus of European water beetles.
In scenario 1, range expansion occurs through successive
bouts of dispersal with subsequent speciation. This would be
generally equivalent to stepping-stone colonization (‘island
hopping’, MacArthur & Wilson, 1967), or, to some extent, to
the progression rule of Hennig (1966). The starting situation is
a small range to which new areas are added sequentially, to be
eventually removed again owing to speciation. The resulting
pattern will be a general positive relationship between phylo-
genetic and geographic distances, with more distant species
having the oldest divergences. This relationship will be
asymmetrical (triangular in a bivariate plot, Fig. 1a): while
there could not be species that are geographically distant but
phylogenetically close (unless there is long-range dispersal),
there could be species that are geographically close but
phylogenetically distant (e.g. species resulting from the initial,
most ancient splits). The age of the species will generally
increase towards the geographic origin of the range expansion
[as postulated by Hennig’s (1966) progression rule]. Typical
examples would be the colonization of archipelagos (Gillespie
& Roderick, 2002; Keppel et al., 2009), or of new available
areas, for example by progressive deglaciation (Hewitt, 2000).
Alternatively, range expansion may occur with subsequent
speciation owing to a fragmentation of the initial range. In this
case the starting situation is the maximum range, which
becomes fragmented and reduced with time, leading to
speciation. Depending on the nature of the barriers fragment-
ing the initial range, one of two outcomes may occur,
providing our second and third scenarios. In scenario 2, if
the range is fragmented as a result of vicariance events that
have a location independent of the distribution of the species,
there should be no correlation between geographic and
phylogenetic distances (Fig. 1b). Typical examples could be
fragmentation of a range by an increase of sea level, or by
tectonic fragmentation of microplates (e.g. Sanmartın, 2003).
Under this scenario, species that are close geographically may
have large phylogenetic divergences, and vice versa. In scenario
3, the range may be fragmented owing to a reduction of gene
flow when there is a progressive and more or less uniform
degradation of the general conditions that allowed the initial
range expansion. This would be equivalent to the refuge
speciation of Moritz et al. (2000) or to the vicariance by niche
conservatism of Wiens (2004), when the barriers resulting
A
4
321
E
D
C
B
A EDCB
4
321
A EDCB
43
21
A
4
321 E
D
C
B
A 43
2
1
E
D
C
B
A EDCB
4 32
1
P
P
P
G
G
G
(a)
(c)
(b)
Figure 1 Schematic representation of the various hypothesized
scenarios of speciation after a range expansion. The first column
shows the geographic distribution of the species (A to E); the
second column, the phylogenetic relationships among them; the
third column, a bivariate plot of the 10 geographic linear distances
(G) versus phylogenetic distances (P) (approximate values; note
that some of the species pairs have identical values). (a) Speciation
by stepping-stone colonization; (b) speciation by vicariance owing
to the formation of random barriers to gene flow (represented by
lines); (c) speciation by increased reduction of gene flow between
favourable patches. The three scenarios result in different patterns
in the relationship between phylogenetic and geographic distances:
(a) triangular relationship, with no closely related species at large
geographic distances; (b) no relationship; and (c) positive rela-
tionship, with neither closely related species at large geographic
distances nor distant phylogenetic species at close geographic
distances (see text for details). Numbers indicate the temporal
sequence of the cladogenetic processes.
The geography of speciation in the Haenydra lineage
Journal of Biogeography 38, 502–516 503ª 2010 Blackwell Publishing Ltd
from suboptimal conditions are not established independently
of the spatial distribution of the species (as in the previous
scenario) but occur through the lines of minimum gene flow,
which would correspond to the largest area with the lowest
population density (Fig. 1c). In this case, the general relation-
ship between geographic and phylogenetic distances should
increase monotonically. The age of the species will generally
increase with respect to the centroid of the distribution of the
clade. Under a perfect scenario, the match between geographic
and phylogenetic distances should be optimal; that is, the
topology of the observed phylogeny should be the one that
minimizes the geographic distances between species.
We test these three scenarios using a lineage of aquatic
beetles with an abundance of narrowly distributed allopatric
species, the ‘Haenydra’ lineage of the genus Hydraena, family
Hydraenidae (Hansen, 1998). Our specific aim is to ascertain
whether the south-western European species of the group can
be said to have persisted in the areas in which they are
currently found since their origin; that is, whether they are
local endemics or have suffered range shifts (e.g. as conse-
quence of glacial cycles) large enough to have erased the
phylogenetic signal of their original distribution. The three
possibilities outlined above are model situations that differ in
the resulting relationship between geographic and phyloge-
netic distances, although the power of the conclusions
depends on the observed pattern: if a strict positive
relationship is found (according to the third scenario), this
can be taken as compelling evidence of a non-random
geographic arrangement of the species. However, the exis-
tence of deviations from this strict proportionality (either
partially, as in Fig. 1a, or completely, as in Fig. 1b) could
indicate either a compliance with the predictions derived
from the initial hypotheses or the breakdown of the expected
pattern owing to subsequent changes in the range of some of
the species.
MATERIALS AND METHODS
Background on the taxonomy of the group
The genus Hydraena Kugelann, with c. 850 known species and
many more to be described, is the most diverse of the family
Hydraenidae, and probably the most diverse among the
aquatic Coleoptera (Jach & Balke, 2008; Perkins, in press).
Adults of most species of Hydraena are typically aquatic, living
in the riparian area of small streams and rivers. Many authors
have recognized groups of apparently closely related species
based on external morphology, which have in some cases been
named as subgenera, or, more recently, as informal species
groups (Jach et al., 2000). One of these traditionally recognized
groups is Haenydra, described by Rey (1886) as a separate
genus, and considered by different authors as a genus (e.g.
Ienistea, 1968; Rocchi, 2009) or a subgenus (e.g. Berthelemy,
1986; Perkins, 1997; Hansen, 1998). In a cladistic analysis of
Hydraena sensu lato, Jach et al. (2000), even acknowledging
their likely monophyly, synonymized Haenydra (together with
all previously defined subgenera with the exception of
Hydraenopsis Janssens), as they would render Hydraena sensu
stricto paraphyletic, and considered it only as an informal
species group (the Hydraena gracilis group).
Currently there are 86 recognized species of the ‘Haenydra’
lineage (Hansen, 1998; Jach, 2004; Audisio et al., 2009), usually
found in clean, fast-flowing waters, often in mountain streams.
They share the absence of parameres in the aedeagus and a
similar external morphology, typically narrow and elongated
(Fig. 2). They are distributed in Europe and the Middle East,
from Iberia to Iran (Hansen, 1998; Jach, 2004). Many species
of this lineage have very restricted distributions, often limited
to a single valley or mountain system, but there are also some
species with very wide geographic ranges, for example
Hydraena gracilis is present throughout Europe from north
Iberia to the Urals (Jach, 2004).
Taxon sampling
We undertook a comprehensive sampling of the western
European species of ‘Haenydra’, including all species occurring
in the Iberian Peninsula, plus a representation of species from
other Mediterranean areas (see Appendix S1 in the Supporting
Information). Missing species in some of the species groups
could be tentatively placed according to the external mor-
phology and the characteristics of the male genitalia. In total,
we included examples of 39 named species of ‘Haenydra’.
The monophyly of Hydraena is generally accepted (Perkins,
1989; Jach et al., 2000; Beutel et al., 2003), but there is no
agreement on the internal phylogeny and, in particular, on the
phylogenetic position of the ‘Haenydra’ lineage. We sampled as
outgroups a comprehensive selection of species of Hydraena,
including most species groups as defined from morphology
(Jach et al., 2000) (Appendix S1). Trees were rooted in
Adelphydraena, hypothesized to be the sister group of Hydra-
ena (Perkins, 1989).
The taxonomy and nomenclature of Jach (2004) is followed
for the species of ‘Haenydra’, except for H. saga. Hydraena saga
as currently recognized has a disjunct distribution, in the
Pyrenean region in the west, and from the eastern Alps to
Bulgaria in the east (Jach, 2004; Fig. 3). Preliminary molecular
and morphological data strongly suggest that the populations
in these two areas are not closely related (unpublished
observations), and we consider here only the Pyrenean
populations as the ‘H. saga complex’.
DNA extraction and sequencing
Specimens were collected alive in the field and directly killed
and preserved in 96% ethanol. DNA was extracted from whole
specimens by a standard phenol-chloroform extraction or
using the DNeasy Tissue Kit (Qiagen GmbH, Hilden,
Germany). Vouchers and DNA samples are kept in the
collections of the Museo Nacional de Ciencias Naturales
(MNCN, Madrid) and the Institute of Evolutionary Biology
(IBE, Barcelona) (Appendix S1). DNA extraction was
I. Ribera et al.
504 Journal of Biogeography 38, 502–516ª 2010 Blackwell Publishing Ltd
0.2
truncata AI508
testacea AI566
monikae AI455
altamirensis AI425
arenicola AI504
gracilis AI333
saga complex AI481
gaditana AI166
riberai AI568
rugosa AI392
devillei AI288
gavarrensis AI1288
dentipalpis AI492
petila AI465
carniolica AI1049
serricollis AI1093
gracilis AI905
bolivari AI171
unca AI276
leonhardi AI339
barrosi AI954
quilisi AI1085
schuleri AI400
catalonica AI381
catalonica AI350
producta AI402
septemlacuum AI795sinope AI788
atrata AI314
heterogyna AI294
hispanica AI329
Hydraenopsis AI423
gracilis AI332
holhausi AI1025
lapidicola AI292
polita AI369
exasperata AI506
gracilis AI1012
hayashii AI691
manfredjaechi AI313
gracilis AI338
bensae AI293
Adelphydraena AI356
evanescens AI286
anatolica AI802
nigrita AI345
fontana AI708
belgica AI426
zezerensis AI182
miyatakei AI828
palawanensis AH133
bitruncata AI354
servilia AI274
tarvisina AI967
morio AI1203
palustris AI309
saga complex AI479
bisulcata AI172
dochula AI518
exarata AI169
gracilis AI510
madronensis AI424
subacuminata AI305
dentipes AI361
polita AI165
truncata AI357
iberica AI181
reyi AI320
tatii AI164
kasyi AI1026
marcosae AI904
capta AI167
lusitana AI385
larissae AI303
devincta AI966
delia AI1061
emarginata AI325
subintegra AH147
Hydraenopsis AI709
monstruosipes AI439
hernandoi AI435
alpicola AI347
integra AI783
catalonica AI1060
balearica AI175
bitruncata AI380
heterogyna AI1051
Hydraenopsis AI456
excisa gr AI391excisa AI348
Hydraenopsis AF90
pygmaea AI346
antiatlantica AI567
corinna AI284
0.87/-
1/100
0.8/55
1/95
0.95/<50
0.95/84
1/99
1/100
0.91/<50
0.77/67
0.89/54
1/78
0.57/<50
0.91/<50
1/100
1/99
1/100
1/100
0.97/54
0.87/<50
1/70
1/92
1/82
1/100
1/100
1/75
1/99
1/<50
1/81
0.74/-
0.84/62
1/100
1/100
1/100
0.99/90
1/100
1/100
1/100
0.64/<50
0.78
1/93
1/100
0.95/<50
0.78/-
1/98
1/100
0.60/<50
1/100
1/1000.8/88
1/100
1/96
0.87/65
0.96/<50
1/53
1/97
1/90
1/90
1/100
0.68/<50
0.82/<50
1/98
H. gracilis lineage
H. iberica lineage
H. dentipes lineage
“Haenydra”
Hydraena
0.53/<500.77/78
0.52/81
1/910.96/730.97/81
1/83
1/93
0.92/83
1/99
0.95/<50
H. iberica clade
H. emarginata clade
H. bitruncata clade
H. tatii clade
Hydraena s.str.
Hydraenopsis
“Phothydraena”
H. rugosa + circulata groups
Figure 2 Phylogram of the species of Hydraena obtained in MrBayes. Numbers at nodes denote Bayesian posterior probability/bootstrap
support in RAxML; ‘-’ marks nodes not present in the RAxML analyses; vertical bars denote the four clades used in the geographic analyses.
Habitus: H. catalonica. See Appendix S1 for the codes of the species.
The geography of speciation in the Haenydra lineage
Journal of Biogeography 38, 502–516 505ª 2010 Blackwell Publishing Ltd
non-destructive, in order to preserve voucher specimens for
subsequent morphometric and morphological study. Typically,
only males were sequenced, and the male genitalia (used for
the identification of the species) were dissected and mounted
prior to the extraction to ensure a correct identification.
We sequenced four fragments, namely two mitochondrial
fragments including four genes (the 3¢ end of cytochrome c
oxidase subunit 1, COI; and the 3¢ end of the large ribosomal
unit plus the leucine transfer plus the 5¢ end of NADH
dehydrogenase subunit 1, rrnL+trnL+nad1) and two nuclear
fragments (small ribosomal unit, SSU; large ribosomal unit,
LSU) (see Appendix S2 for the primers used). For each
fragment both forward and reverse sequences were obtained.
In some specimens the COI fragment was amplified using
internal primers to obtain two fragments of around 400 bp
each (Appendix S2). Sequences were assembled and edited
with Sequencher 4.7 (Gene Codes, Inc., Ann Arbor, MI).
New sequences have been deposited in GenBank with
accession numbers HM588308–HM588600 (Appendix S1).
Protein-coding genes were not length-variable, and the
ribosomal genes were aligned with the online version of
mafft 6 using the G-INS-i algorithm and default parameters
(Katoh & Toh, 2008).
Phylogenetic analyses
Bayesian analyses were conducted on a combined data matrix
with MrBayes 3.1.2 (Huelsenbeck & Ronquist, 2001), using
five partitions corresponding to the sequenced genes (the
rrnL+trnL fragment was considered a single partition) and a
GTR+I+G model independently estimated for each partition.
MrBayes was run for 15 · 106 generations using default
values, saving trees after every 500 generations. ‘Burn-in’
values were established after visual examination of a plot of the
standard deviation of the split frequencies between two
simultaneous runs.
We also used maximum likelihood as implemented in the
on-line version of RAxML (which includes an estimation of
bootstrap node support, Stamatakis et al., 2008), using
GTR+G as the evolutionary model and the same five gene
partitions as used in MrBayes.
Estimation of divergence times
To estimate the relative age of divergence of the lineages we
used the Bayesian relaxed phylogenetic approach implemented
in beast 1.4.7 (Drummond & Rambaut, 2007), which allows
(a) (b)
(c) (d)
Figure 3 Maps with the distribution of the species of Hydraena of the ‘Haenydra’ lineage included in the clades used in the geographic
analyses, with their centroids (see Table 1 for the coordinates). (a) Hydraena iberica clade; (b) H. emarginata clade; (c) H. bitruncata clade;
(d) H. tatii clade (in the Alpes Maritimes, minimum distance from the species of the Alpine group).
I. Ribera et al.
506 Journal of Biogeography 38, 502–516ª 2010 Blackwell Publishing Ltd
variation in substitution rates among branches. We imple-
mented a GTR+I+G model of DNA substitution with four rate
categories, using the mitochondrial data set only and pruning
specimens with missing gene fragments. We used an uncorre-
lated lognormal relaxed molecular clock model to estimate
substitution rates and the Yule process of speciation as the tree
prior. Well-supported nodes in the analyses of the combined
sequence (mitochondrial and nuclear) were constrained to
ensure that the beast analyses obtained the same topology. We
ran two independent analyses for each group, sampling every
1000 generations, and used Tracer 1.4 to determine conver-
gence, measure the effective sample size of each parameter and
calculate the mean and 95% highest posterior density interval
for divergence times. The results of the two runs were combined
with LogCombiner 1.4.7, and the consensus tree was compiled
with TreeAnnotator 1.4.7 (Drummond & Rambaut, 2007).
The analyses were run for 30 · 106 generations, with the
initial 10% discarded as burn-in. Because of the absence of a
fossil record with which to calibrate the trees we used as a prior
a rate of 2.0% of pairwise divergence per million years
(Myr)1), established for subterranean species of a closely
related family (Leiodidae) for a combination of mitochondrial
markers (including those used here) and using as a calibration
point the tectonic separation of the Sardinian microplate
(Ribera et al., 2010a). We set as a prior rate a normal
distribution with an average rate of 0.01 substitutions
site)1 Myr)1, with a standard deviation of 0.001.
Geographic analyses
Contour maps of the distributions of the species of the
‘Haenydra’ lineage included in the studied clades were
compiled from published and unpublished sources (Jach,
2004; Sanchez-Fernandez et al., 2008; Checklist of the species
of the Italian fauna, v. 2.0, http://www.faunaitalia.it) (Fig. 3).
Species’ range centroids (centre of mass of the polygon
representing the distribution of a species) and distances
between centroids were calculated using ArcGIS 9.2 (Envi-
ronmental Systems Research Institute Inc., Redlands, CA,
USA) (Table 1). To check the association between phyloge-
netic and geographic distances between the centroids we used
the following three approaches.
1. Bivariate plots of the linear distance between the centroids of
the species ranges and the branch lengths of the ultrametric
trees, which is the estimated age of divergence between species
(i.e. their phylogenetic distance) (Fig. 1).
2. Mantel tests for the general association between the matrices
of geographic and phylogenetic distances. Multiple Mantel
tests were performed using zt 1.1 (Bonnet & Van de Peer,
2002), with 10,000 randomizations to generate a null distri-
bution and assess the significance of the results.
3. An optimization procedure to assess the match between the
observed geographic distribution and the topology obtained
from the phylogeny. We first compute the Euclidean minimum
spanning tree (EMST), that is, the minimum spanning tree of a
set of n points in the plane (the centroids of the distributions),
where the weight of the edge between each pair of points is the
linear distance between those two points. The EMST connect-
ing n vertices will have n (n ) 1)/2 edges, which are computed
through a standard minimum spanning tree algorithm (see e.g.
Graham & Hell, 1985), a trivial task for graphs of fewer than six
nodes. The result is a graph connecting all points minimizing
the weight of the edges, that is, the distances among centroids.
In the scenario outlined in Fig. 1c, the temporal sequence of
cladogenetic events will be determined by the length of the
edges connecting the centroids: the first split will be between the
taxa at either extreme of the longest edge, the second will be
between taxa at either end of the second longest, and so on until
the two closest species are separated.
To assess the probability that the observed topology could
be identical to that obtained with this optimization procedure,
we obtained all possible unrooted bifurcating topologies of
each of the studied clades in paup* 2 (Swofford, 2002), and
considered them as a null distribution against which the
probability of each individual topology was estimated. Note
that the use of the EMST determines not only the final
topology but also the relative order of all the cladogenetic
events. We did not consider the relative order in the cases in
which the cladogenesis occurs in two different branches of the
tree, as this does not affect the final topology.
RESULTS
Phylogeny of Hydraena
The final matrix included 94 taxa and 2831 aligned characters.
Part of the rrnL+trnL+nad1 fragment was missing for two
Table 1 Centroids of the distributions of the species of Hydraena
in the clades used for the geographic analyses (in decimal coor-
dinates). See Fig. 3 for the distributions of the species.
Clade Species X Y
H. iberica H. altamirensis )4.868 39.504
H. iberica )6.474 41.221
H. lusitana )7.614 41.277
H. madronensis )4.316 38.391
H. emarginata H. emarginata )2.746 42.894
H. hispanica )7.157 41.588
H. larissae 9.395 46.041
H. saga complex )0.015 42.630
H. tarvisina 11.588 46.003
H. bitruncata H. bensae 6.000 44.175
H. bicuspidata 4.430 45.350
H. bitruncata 2.314 42.253
H. catalonica p 1.262 42.415
H. catalonica m 2.384 41.815
H. polita 6.071 45.869
H. tatii H. gaditana )5.293 36.679
H. manfredjaechi )2.541 38.228
H. monstruosipes )6.858 42.771
H. tatii )3.757 37.132
H. zezerensis )7.584 40.385
The geography of speciation in the Haenydra lineage
Journal of Biogeography 38, 502–516 507ª 2010 Blackwell Publishing Ltd
species (Appendix S1), and for two of the repeated specimens
of H. gracilis only COI was sequenced. The nuclear markers
(SSU and LSU) were sequenced only for a representation of the
species of ‘Haenydra’ owing to the general low variability
within this lineage, with many identical sequences between
closely related species (Appendix S1).
The runs of MrBayes converged to split frequencies lower
than 0.01 at 11 · 106 generations, leaving a total of 4 · 2 · 106
generations for the sampling of the tree space (= 16,000 trees).
The monophyly of Hydraena and the basal relationships
among its major clades were not well supported (Fig. 2). There
are five well-supported lineages within the genus Hydraena: (1)
the subgenus Hydraenopsis (as defined in Jach et al., 2000); (2)
the South African H. monikae; (3) the ‘Phothydraena’ lineage
(H. testacea species group in Jach et al., 2000); (4) the
H. rugosa and H. circulata species groups, sisters with good
support in the analysis with Bayesian probabilities although
not in the maximum likelihood analysis; and (5) the main
lineage within Hydraena sensu stricto (including the ‘Haenydra’
lineage), which was well supported in both analyses (Fig. 2; see
Appendix S1 for the composition of the species groups).
Within the main lineage of Hydraena sensu stricto, the
H. palustris group was sister to the rest (in agreement with
Jach et al., 2000), which were in turn divided into three well-
supported main clades: (1) H. bisulcata and its allies; (2) a
clade broadly including the H. riparia, H. minutissima,
H. rufipes and H. holdhausi groups; and (3) the ‘Haenydra’
lineage (Fig. 2). The relationship between these three main
clades was not well resolved, with MrBayes favouring a sister
relationship between ‘Haenydra’ and the H. bisulcata group,
and RAxML favouring a sister relationship with the H. riparia
group (sensu lato), albeit in both cases with low support. In all
cases the monophyly of the ‘Haenydra’ lineage was strongly
supported (Bayesian posterior probability, BPP = 1.0; maxi-
mum likelihood bootstrap, MLb = 100%; Fig. 2).
Phylogeny of the ‘Haenydra’ lineage
There were three well-supported main lineages within ‘Hae-
nydra’, namely the Hydraena iberica, H. gracilis and H. dentipes
lineages (Fig. 2). There were also three species with an isolated
position, namely Hydraena carniolica, H. schuleri and H. sub-
integra.
The H. iberica lineage included four Iberian endemics
(Fig. 3a). The H. gracilis lineage included the H. emarginata
clade, with the Iberian endemics H. saga complex, H. emargi-
nata and H. hispanica as sister to two species from the Alps
(Figs 2 & 3).
The third main group within ‘Haenydra’, the H. dentipes
lineage, included two clades with narrow-range western Med-
iterranean endemics. The first, the Hydraena bitruncata clade,
included H. catalonica, H. bitruncata, H. polita and H. bensae
(Figs 2 & 3c). The first two species have narrow distributions in
the north-east of the Iberian Peninsula and southern France, H.
polita has a widespread distribution from north Iberia to the
eastern Alps, and H. bensae is endemic to the Alpes Maritimes
(Fig. 3c). The sister of this clade was not well established. All of
these species lack a small flagellum in the apical part of the
median lobe of the aedeagus, which is present in the rest of the
species of the H. polita group (H. dentipes, H. producta and
H. heterogyna among those included in the study). Hydraena
bicuspidata, from the Massif Central in south-east France (close
to Lyon, Ganglbauer, 1901), should also be included in this
clade, as it lacks the flagellum and has a very similar body shape.
The second clade within the H. dentipes lineage was the
H. tatii clade, including five Iberian narrow-range endemics,
H. tatii, H. manfredjaechi, H. gaditana, H. zezerensis and
H. monstruosipes, as sister to some species in the Alps and Italy,
namely H. devincta, H. devillei, H. leonhardi and H. lapidicola
(Figs 2 & 3d). The sister relationship between the species
H. tatii, H. manfredjaechi and H. gaditana (i.e. the ‘H. tatii
group’ sensu Fresneda et al., 1994) with H. zezerensis plus
H. monstruosipes was not well supported, although the node
was present in all analyses (maximum likelihood and Bayesian,
both with the full combined sequence and with the mitochon-
drial data only). The sister group of the H. tatii clade was
H. truncata (although with low support, Fig. 2), which has a
widespread European distribution that includes the north-west
of the Iberian Peninsula. There are three likely missing species
in this clade: H. sanfilippoi, close to H. lapidicola (Audisio & De
Biase, 1995); H. bosnica, close to H. leonhardi (Audisio et al.,
1996); and H. hungarica, also related to H. bosnica and
H. leonhardi (the three share with other species of the group
the female gonocoxite with two small symmetric depressions).
Hydraena sappho Janssens, from the small island of Levkas
(Greece), has been associated with the H. tatii clade (Audisio
et al., 1996). A closer examination of the only known specimen
(the holotype, Janssens, 1965) showed that it is most likely to
be related to species from the eastern Mediterranean, not to
the Iberian species (M.A. Jach, unpublished observation).
Two of the species of ‘Haenydra’ were found to be
paraphyletic: in the H. gracilis complex (sensu Jach, 1995),
the north Iberian populations were sister to specimens sampled
from sites from Britain to Turkey, including H. anatolica (Jach,
1995); and specimens of H. catalonica from the Montseny
Massif (central Catalonia) were sister to the Pyrenean
H. catalonica plus H. bitruncata. For the geographic analyses,
H. catalonica was split into its two geographic lineages, the
populations from the Pyrenees (‘H. catalonica p’) and the
populations from the Montseny Massif (‘H. catalonica m’).
Estimation of divergence times
From the results of the beast runs, and using a calibration of
0.01 substitutions site)1 Myr)1, the origin of the ‘Haenydra’
lineage was estimated to be c. 8.5 Ma (late Miocene), with a wide
confidence interval (Fig. 4). The three main lineages (H. iberica,
H. gracilis and H. dentipes lineages) originated c. 6 Ma, and most
species, including all Iberian endemics, were estimated to be less
than 2.6 Myr old, that is, of Pleistocene origin (Fig. 4). There
were relatively deep divergences within some of the species
with wider distributions, such as H. polita (0.8 Myr between
I. Ribera et al.
508 Journal of Biogeography 38, 502–516ª 2010 Blackwell Publishing Ltd
specimens from the Pyrenees and south Germany) and
H. heterogyna (0.8 Myr between specimens from central Italy
and the French Alpes Maritimes; Fig. 4; Appendix S1), suggest-
ing the possible existence of unrecognized cryptic diversity.
Geographic analyses
For the geographic analyses, four well-supported clades of
‘Haenydra’ were selected, including most of the south-western
narrowly distributed species and for which the sampling
(according to morphology) could be considered complete or
with at most one or two missing species. These were the
H. iberica (1), H. emarginata (2), H. bitruncata (3) and H. tatii
(4) clades (see above and Fig. 2 for their composition, and
Fig. 3 for the distribution of the species).
The H. iberica lineage had only four species, which is below
the minimum number necessary for the implementation of
Mantel tests in zt (Bonnet & Van de Peer, 2002). For the other
1.0 Myr
lapidicola AI292
altamirensis AI425
bensae AI293
emarginata AI325
polita AI165
gracilis AI333
gracilis AI510
monstruosipes AI439
producta AI402
excisa AI348
excisa gr AI391
gaditana AI166
devincta AI966
manfredjaechi AI313
dentipes AI361
tarvisina AI967
madronensis AI424
gracilis AI338
saga complex AI481
iberica AI181
gracilis AI332
saga complex AI479
belgica AI426
lusitana AI385
septemlacuum AI795
catalonica p AI381
heterogyna AI1051
schuleri AI400
catalonica p AI1060
truncata AI508
heterogyna AI294
gracilis AI905
polita AI369
anatolica AI802
leonhardi AI339
bitruncata AI380
catalonica m AI350
integra AI783
tatii AI164
alpicola AI347
devillei AI288
hispanica AI329
larissae AI303
zezerensis AI182
bitruncata AI354
carniolica AI1049
sinope AI788
truncata AI357
gracilis AI1012
subintegra AH147
exasperata AI506
evanescens AI286
1.9
0.14
1.1
0.06
0.15
5.1
0.5
1.8
0.05
8.5
6.1
1.0
0.08
1.4
5.5
1.7
3.5
0.9
0.4
0.3
5.9
2.7
0.3
0.12
0.8
3.7
4.5
1.3
0.6
0.2
0.14
0.4
0.7
0.14
2.3
2.37.0
0.9
3.3
1.2
0.3
0.5
6.1
7.6
0.8
5.2
1.6
0.8
5.0
2.4
3.1
H. gracilis lineage
H. iberica lineage
H. dentipes lineage
H. tatii clade
H. bitruncata clade
H. iberica clade
00
0
0 0
0
00
0
0
0
0
0
0
0
0
0
0
00
0
00
0
0
0
H. emarginata clade
Figure 4 Ultrametric tree obtained using beast, using the mitochondrial data from a selection of the Hydraena specimens in the
‘Haenydra’ lineage and constraining the well-supported nodes of the topology represented in Fig. 2 (black circles). The four clades studied in
the geographic analyses are marked in red. Numbers at nodes denote the age estimate (Ma); blue bars, 95% confidence intervals; vertical bar,
the Pliocene/Pleistocene transition (2.6 Ma). See Appendix S1 for the codes of the species.
The geography of speciation in the Haenydra lineage
Journal of Biogeography 38, 502–516 509ª 2010 Blackwell Publishing Ltd
clades, geographic and phylogenetic distances, as measured
with a Mantel test with 10,000 permutations, were significantly
correlated in the H. emarginata (r = 0.9, P < 0.05) and the
H. tatii clades, both when only the five Iberian species were
included (r = 0.90, P < 0.05) and when the pooled Alpine
species were included as a non-overlapping outgroup to the
Iberian species (r = 0.83, P < 0.01) (Table 2). In the
H. bitruncata clade, the Mantel test was not significant at the
standard level (Table 2; r = 0.68, P = 0.08). When the missing
species H. bicuspidata was added to the phylogeny as sister to
H. bitruncata plus H. catalonica (the most likely position
according to morphology, see above), it became significant,
despite the reduction in the correlation, owing to the increase
in power (r = 0.32; P < 0.05).
The bivariate plot between geographic and phylogenetic
distance for the H. iberica clade had three data points clearly
outside a hypothetical linear relationship, corresponding to the
pairwise distances including H. iberica (Fig. 5a). The centroid
of the distribution of this species was too close to H. lusitana
and too distant from H. altamirensis and H. madronensis to
obtain a linear relationship between geographic and phyloge-
netic distance, suggesting a possible secondary range expansion
of H. iberica. To explore this possibility, we sequenced the COI
fragment of five additional specimens of H. iberica through
their range (Appendix S1). They had identical sequences, with
the exception of a difference of one base pair for the specimen
from south Portugal (voucher MNCN-AI386, Appendix S1),
at the south-western limit of the species distribution (Fig. 3a),
supporting the hypothesis of a recent expansion.
For the H. emarginata and H. tatii clades, the bivariate plot
(Fig. 5b, d) showed a monotonic increase of phylogenetic
distance as the distance between centroids increased, without
apparent outliers and in particular without phylogenetically
distant species pairs in close geographic proximity. In the
H. bitruncata clade, the pairwise distances corresponding to
H. polita (the more widespread species of the group) were
outliers from a linear relationship in the bivariate plot
(Fig. 5c).
For the H. iberica and H. bitruncata clades, the observed
topology was not in agreement with that obtained from the
EMST (Fig. 6a, c). The topology optimizing the geographic
distances (EMST) for the H. iberica clade placed H. altamir-
ensis as sister to H. madronensis, in contrast with the observed
relationship (H. iberica sister to H. madronensis, Fig. 2). In any
case, owing to the low number (three) of different unrooted
trees for four taxa, the observed tree could not be said to be
different from a random geographic arrangement. For the
H. bitruncata clade, two topologies had a better match with the
EMST: [(H. bitruncata, H. catalonica m) H. catalonica p] and
[(H. catalonica m, H. catalonica p) H. bitruncata] (Fig. 6c,
Table 2 Matrices of the geographic linear distances between the centroids of the species of Hydraena (in km) and the phylogenetic distances
(i.e. age estimates, in Ma). See text and Fig. 2 for the sister of the H. tatii clade. In the H. bitruncata clade, ‘H. catalonica p’ refers to
populations from the Pyrenees, and ‘H. catalonica m’ to populations from the Montseny Massif (central Catalonia). In the H. bitruncata
clade, the age estimate of H. bicuspidata (not included in the study) is the middle point of the branch between the two nodes in which it is
hypothesized to be placed based on morphological evidence (see text).
H. iberica clade H. altamirensis H. iberica H. lusitana
H. iberica 236 km/1.6 Ma
H. lusitana 308/4.5 96/4.5
H. madronensis 133/2.3 367/2.3 431/4.5
H. emarginata clade H. emarginata H. hispanica H. larissae H. saga complex
H. hispanica 390/1.0
H. larissae 1025/2.0 1415/2.0
H. saga complex 226/0.6 601/1.0 840/2.0
H. tarvisina 1190/2.0 1580/2.0 170/0.7 998/2.0
H. bitruncata clade H. bensae H. bicuspidata H. bitruncata H. catalonica p H. catalonica m
H. bicuspidata 175/4.8
H. bitruncata 367/4.8 380/1.9
H. catalonica p 431/4.8 415/1.9 89/0.1
H. catalonica m 394/4.8 450/1.9 49/1.8 115/1.8
H. polita 188/4.8 100/3.8 501/3.8 542/3.8 538/3.8
H. tatii clade H. gaditana H. manfredjaechi H. monstruosipes H. tatii H. zezerensis
H. manfredjaechi 296/3.2
H. monstruosipes 691/6.5 628/6.5
H. tatii 146/1.0 161/3.2 683/6.5
H. zezerensis 460/6.5 501/6.5 271/2.5 496/6.5
[sister outgroup] 1300/7.5 1000/7.5 1100/7.5 1100/7.5 1300/7.5
I. Ribera et al.
510 Journal of Biogeography 38, 502–516ª 2010 Blackwell Publishing Ltd
P = 3/15 = 0.2). With the inclusion of H. bicuspidata in its
hypothesized phylogenetic position there were five topologies
that matched the EMST better than the observed one, with a
marginal significance (P = 6/105 = 0.057).
For the H. emarginata and H. tatii clades, of all possible
unrooted topologies with five taxa (15), the observed one was
identical to that determined by the EMST. The observed
relative order of two of the nodes in different branches in each
m
l i
a
123
1
2
3
he
l t
s
1
4
2 3
23
1 4
3
1
2
3
2
1
5
2
1
ma
t
z
mo
g
1
4
2
3
monstruosipes zezerensis
gaditana tatii
manfredjaechi
[outgroup]
3
4
larissae tarvisina
emarginata saga complex
hispanica
3
4
lusitana madronensis
iberica altamirensis
bensae polita
catalonica p bitruncata
catalonica m
lusitana
madronensis
iberica altamirensis
bensae polita
catalonica m bitruncata
catalonica p
o
5
p
cp bi
be
cm
1
43
21
4 3
2
1
43
2
3
4
4
3
4
31
3
2
4
(a)
(c) (d)
(b)
Figure 6 Schematic representation of the Euclidean minimum spanning trees (EMSTs) among the centroids of the species in the four
clades of the ‘Haenydra’ lineage used for the geographic analyses: (a) Hydraena iberica clade; (b) H. emarginata clade; (c) H. bitruncata clade;
(d) H. tatii clade. Circles represent the geographic positions of the centroids, as in Fig. 3. Lines between centroids represent cladogenetic
events. Numbers represent the temporal order of the cladogenetic events. In (b) and (c), the reconstructed phylogeny (taken from Fig. 4)
agrees with the EMST, except for the temporal order of some nodes in various branches, which does not affect the topology (numbers in
blue, observed sequence; in red, temporal sequence according to the EMST). In (a) and (c) the observed phylogeny (in blue) does not agree
with the phylogeny reconstructed from the EMST (in red).
(b)(a)
(c) (d)
5
0
1
2
3
4
0
1
2
3
4
5
6
0
1
2
0
1
2
3
4
5
6
7
8
100 500400300200
600500400300200100
200015001000500
15001000500
GG
G G
P P
PP
Figure 5 Bivariate plots of the geographic
(linear) distance between the centroids of the
‘Haenydra’ species (G, km) versus the phy-
logenetic distance (P, age estimate in Ma). (a)
Hydraena iberica clade; grey circles, distances
to H. iberica; (b) H. emarginata clade; (c)
H. bitruncata clade; grey circles, distances to
H. polita; white circles, distances to
H. bicuspidata; (d) H. tatii clade.
The geography of speciation in the Haenydra lineage
Journal of Biogeography 38, 502–516 511ª 2010 Blackwell Publishing Ltd
of the trees was reversed with respect to the order expected
with the EMST (Fig. 6b, d), although they were estimated to
have occurred in close temporal proximity, and the 95%
confidence intervals fully overlap each other (Fig. 4). Owing to
the low number of possible topologies (15), the geographic
distribution of the species could only be considered to be
marginally different from random (P < 1/15 = 0.067). For the
H. tatii clade, when the fully allopatric Alpine sister group was
included with a pooled geographic distance considered to be
the shortest from the Iberian Peninsula (i.e. the Alpes
Maritimes, Table 2), the relationship became highly significant
owing to the increase to six taxa (P = 1/105 < 0.01). Similarly,
alternative topologies to the sister relationship between the
H. tatii group and H. monstruosipes plus H. zezerensis (placing
each of them as sister to a species pair from the Alps) resulted
in equally significant non-random geographic distributions
when the two groups were considered separately.
DISCUSSION
Origin and phylogeny of the ‘Haenydra’ lineage
There was strong support for the monophyly of the species of
the ‘Haenydra’ lineage, including H. schuleri, which has a
deviating morphology. Although the phylogenetic position of
the lineage was not resolved, it was nested within the main
clade of Hydraena sensu stricto, in agreement with the
conclusions of Jach et al. (2000), and not as sister to the rest
of the genus, as hypothesized by Berthelemy (1986) and
Perkins (1997). Jach et al. (2000) suggested that the
H. armipalpis group (China) could be the sister to Haenydra,
owing to the base of the parameres being fused with the
median lobe of the aedeagus and a similar general structure of
the pronotum and elytra, but no species of this group could be
obtained for the molecular study.
Within the wider Hydraena, the main trends of our
phylogeny also agree with the results of Jach et al. (2000),
with several well-defined lineages including Hydraenopsis and
other species groups considered to be plesiomorphic
(H. monikae, H. rugosa group, ‘Phothydraena’ and H. circulata
group). Our results confirm the inclusion of the species of the
H. minutissima group (the former ‘Hadrenya’) within the main
lineage of Hydraena sensu stricto, as hypothesized by Jach et al.
(2000), but not close to ‘Haenydra’, as assumed by previous
authors (d’Orchymont, 1925; Perkins, 1997).
According to our calibration, the ‘Haenydra’ lineage dates
from the late Miocene, but the main diversification of the
group, and the origin of most of the species, took place during
the Pleistocene. Post-Pliocene diversification would explain the
absence of any species of the group in North Africa, despite
intensive search efforts in the area by numerous entomologists
and the obvious dispersal abilities of some species through
continuous landmasses. There is a record of Hydraena exasp-
erata from Morocco in d’Orchymont (1935) (a male, deposited
in the Institut royal des Sciences naturelles de Belgique,
Brussels), but, as suggested by the same author, it may be a
labelling mistake or a specimen carried over from the previous
collecting sites in south Spain. The only species present in
western European islands are H. gracilis in Britain and Ireland,
which are likely to have been connected to mainland Europe
during the early Holocene after the Last Glacial Maximum
(Lambeck & Chappell, 2001), and three species in Corsica and
Sardinia (Audisio et al., 2009). The latter species are often
hypothesized to be the result of vicariance owing to the tectonic
separation of the Corsican microplate during the Oligocene
(see Audisio et al., 2009 for an overview of possible scenarios).
According to our estimations based on the same vicariant split
in a related family (Leiodidae, Ribera et al., 2010a), the
Corsican Hydraena evanescens has an estimated age of
5.2 Ma, that is, the end of the Messinian. This opens the
possibility of a colonization of Corsica and Sardinia through
land connections during the late Miocene, with vicariance of
the island populations after the opening of the Straits of
Gibraltar with the re-filling of the Mediterranean (Garcıa-
Castellanos et al., 2009). Some other Sardinian endemics have
been estimated to be of a similar age, in particular some cave
salamanders of the genus Hydromantes (Carranza et al., 2008),
and the painted frog Discoglosus sardus (Zangari et al., 2006).
The dispersal of these taxa would have required a land corridor
that was probably also suitable for Haenydra.
All the Iberian endemic species were estimated to be of
Pleistocene age, in agreement with some groups of aquatic
beetles (e.g. family Dytiscidae, Ribera & Vogler, 2004), but in
sharp contrast with others, such as the two Iberian endemic
species of Enicocerus Stephens (Hydraenidae, Ribera et al.,
2010b) and the western Mediterranean species of Hydrochus
(A. Hidalgo-Galiana & I. Ribera, in preparation), all of them of
late Miocene origin. The most widespread and common
species of the lineage, H. gracilis, has a recent origin nested
within a clade with predominantly narrow-range endemics.
Differences between specimens through the range (Latvia,
Britain, Bulgaria) were minimal, strongly suggesting a recent,
late Pleistocene range expansion. By contrast, other widespread
species of the group (H. polita, H. truncata) have deep
divergences between specimens from different parts of their
species’ range (estimated to be c. 1 Myr), suggesting the
existence of frequent cryptic diversity within the lineage. This
is also the case for the species found to be paraphyletic
(including H. gracilis sensu lato), or known to be a composite
of polyphyletic lineages (as for the current concept of H. saga).
Geographic analyses
The difficulty in determining past geographic ranges in the
absence of a fossil record is a major impediment in studies of
speciation and diversification (Gaston, 2003; Losos & Glor,
2003). In some cases the accumulation of indirect evidence
(genetic, ecological, geographic) adds to the credibility of a
given scenario, but it is often not possible to contrast its
likelihood against alternative possibilities. The use of null
models as a reference for the comparison of the observed
pattern allows a more rigorous assessment (Barraclough &
I. Ribera et al.
512 Journal of Biogeography 38, 502–516ª 2010 Blackwell Publishing Ltd
Nee, 2001). We did not test the geography of speciation in the
classic sense (Mayr, 1963), because this is not possible using
only data of current distributions. In our scenario, speciation
ultimately may occur as a result of isolation through rarefac-
tion of the range, but whether this can be considered allopatric
or peripatric depends on the precise distribution of the species
at the time of speciation, which is not possible to establish
without detailed genetic data (Butlin et al., 2008).
The finding that for a given clade there is a strong
correlation between geographic and phylogenetic distance is
evidence of the non-randomness of the spatial distribution of
the species. This correlation may be the result of a process in
which the probability of speciation is inversely proportional to
the distance to the nearest neighbours, as could happen in the
progressive rarefaction of a species’ range owing to changes in
general climatic or ecological conditions (Fig. 1c). In the
traditional models of isolation by distance (Wright, 1943;
Kimura, 1953), geographic distance is also proportional to
genetic distance, but this proportionality is maintained
through the continuous presence of gene flow, more likely to
occur between neighbouring individuals (or populations).
Given a widespread species showing isolation by distance
through its range, if the general conditions were to deteriorate
so that gene flow diminished progressively, it could be
expected that the first interruption would occur among the
groups of populations separated by the longest distance,
followed by the next longest, and so on successively, resulting
in a topology matching an EMST among the final species.
We have not assessed the degree of overlap between species
(as in Lynch, 1989 or Barraclough & Vogler, 2000) because
our results depend only on the relative position of the
centroid of the distribution of each species in relation to the
others, not on possible range expansions or contractions
around this centroid. The relative position of the centroids
seems to be less evolutionary labile than the extension of the
range, which will usually change faster than the rate of
speciation (Coyne & Orr, 2004). Our scenarios also do not
assume that the range of the ancestral species is the sum of
that of the descendants, only that the relative position of the
centroid of their distribution is intermediate between the two.
Of the two statistical tests we applied to the geographic
data, the Mantel test is the least stringent, and may still be
significant when the position of the centroids of two close
species swap, or when (as a result of, for example, a
geographically biased range expansion) the current centroid
changes its relative position with respect to the nearest
neighbours. By contrast, the optimization test through the
comparison of the topologies is more restrictive, in that any
change of the relative position of even nearest-neighbour
species would result in a suboptimal topology. This is clearly
exemplified in the H. bitruncata clade, which has a significant
correlation between geographic and phylogenetic distances as
measured with a Mantel test when H. bicuspidata is included
in its most likely position, but is not significantly different
from a random arrangement as there are several topologies
with a better match to the geographic optimum.
A limitation of our approach is that, while a positive result is a
clear indication that species have maintained their relative
geographic positions, when there is no significant correlation,
or the topology does not optimize geographic distances, it is not
possible to affirm that there has been range movement. As seen
in Fig. 1, other modes of speciation (e.g. vicariance by random
breaks) will result in this lack of correlation even if the species
remain in place. Even assuming that the main diversification
mechanism of the ‘Haenydra’ lineage is the succession of cycles
of expansion of some species with subsequent fragmentation
(Fig. 1c), the geographic signal, as measured here, will persist
only until the next expansion of a species of the clade. This could
be the case in the H. iberica clade, in which both the genetic
uniformity and the deviation from the linear correlation of
H. iberica strongly suggest a recent expansion from its original
range, which may be in central Iberia based on the interpolation
of the geographic distances and assuming a linear relationship
with phylogenetic distances in Fig. 5a. The most widespread
species of the ‘Haenydra’ lineage, H. gracilis, was found to be of
recent origin (c. 0.5 Ma), and the few data available show that
the central and northern European form is very homogeneous
through its range, as expected after a recent expansion.
Although we did not analyse this clade in detail owing to the
possibly high number of closely related missing species, the
recognition of distinct taxa in the periphery of its current
distribution (Jach, 1995) suggests that this could be an example
of a species complex in the early stages of range fragmentation.
Some of these peripheral taxa are, however, island endemics
(e.g. Hydraena elisabethae on the island of Thassos, and H. nike
in Samothrace; Jach, 1995). In these cases, the isolation would
be produced by vicariant barriers and there does not need to be
a correlation between geographic and phylogenetic distances.
The H. tatii clade shows the strongest evidence for a non-
random distribution among the four tested. The five Iberian
endemics have a common origin by the late Pliocene. This is
coincident in time with an acute cooling period that may have
facilitated the expansion of the ancestral species, prior to the
origin of the Mediterranean climate at c. 3.1–3.2 Ma, with its
strong seasonality and increase in aridity (Suc, 1984; Mijarra
et al., 2009). Subsequent cladogenetic events within the H. tatii
clade would have taken place during the Pleistocene glacial
cycles within the Iberian Peninsula, and without changes in the
geographic location of the resulting species – or at least with
changes not large enough to erase the geographic signal in their
current distributions.
Although with lower support, the H. bitrunctata clade also
showed some evidence of geographic structure when
H. bicuspidata was included in its most likely phylogenetic
position, with a significant overall correlation between geo-
graphic and phylogenetic distances. The origin of this clade
was estimated to be at the Pliocene–Pleistocene transition,
again a cold period (Lisiecki & Raymo, 2007) that could have
made possible the expansion of species typical of cold
mountain streams. The distribution of this clade, between
north-east Iberia (north of the Ebro valley) and the Alps falls
outside the traditionally recognized Pleistocene refugia (the
The geography of speciation in the Haenydra lineage
Journal of Biogeography 38, 502–516 513ª 2010 Blackwell Publishing Ltd
southern peninsulas), but still shows signs of conservation of the
geographic ranges.
Glacial cycles may induce regular expansions during
favourable times (either glacials or interglacials, depending
on the autecology of the species), followed by range contrac-
tion to refugia when conditions turn adverse (Dynesius &
Jansson, 2000; Bennett & Provan, 2008; Stewart et al., 2010).
These repeated cycles produce different degrees and patterns of
phylogeographic structure (Hewitt, 2000). However, it seems
that for some lineages, among them Haenydra, the range
expansions are infrequent and affect only some taxa. Periods of
range contraction result in the generation of multiple, isolated
residual species. The process would thus not be cyclical, in the
sense that conditions do not return to the same original state,
but accumulative: each expansion produces a set of new species
that do not contribute to the next cycle, and overlap with the
species resulting from the previous ones. The concept of
‘refugia’ (as defined by Stewart et al., 2010) will apply not to
species, as they would not suffer cyclical periods of contrac-
tion–expansion, but to the lineage: successive glaciations
would eradicate populations (or species) in the glaciated areas,
allowing the survival (and accumulation) of the species only in
the refugia. If the species are able to expand their ranges only
occasionally, as seems to be the case for ‘Haenydra’, either they
remain in the refugia as narrow-range endemics, or, when they
expand, they do not mix with the populations that are left, as
they would already be different species.
Concluding remarks
With our approach we have shown that under some circum-
stances it is possible to obtain strong evidence of stasis of the
geographic ranges of narrow-range endemic species through
the study of their phylogenetic relationships and their current
distributions. At least some of the studied clades seem likely to
be formed by true endemics, originating in the areas in which
they are currently found through the fragmentation of a more
widely distributed species during the late Pliocene and
Pleistocene. This speciation within refugia supports increasing
evidence of the complexity of the evolutionary processes that
took place in the Mediterranean peninsulas during Pleistocene
glacial cycles, with refugia being far more than simple
repositories of accumulated genetic diversity (Gomez & Lunt,
2007).
ACKNOWLEDGEMENTS
We thank all colleagues mentioned in Appendix S1 for supply-
ing material for study, A. Faille and A. Hidalgo for help in the
laboratory, W. Zelenka for the habitus of Fig. 2, P. Abellan for
geographic data, and A. Cieslak, J. Castresana, P. Abellan and
two anonymous referees for comments. This work was funded
through the EU program Synthesys (projects AT-TAF-217,
1613, 2201 and 2391 to I.R., J.A.D., L.F.V. and J.G., respectively)
and projects CGL2004-00028 and CGL2007-61665 to I.R.
REFERENCES
Audisio, P. & De Biase, A. (1995) Materials towards a bio-
geographical analysis of Italian Hydraena and Haenydra
(Coleoptera, Hydraenidae). Fragmenta Entomologica, 27,
163–189.
Audisio, P., De Biase, A. & Jach, M.A. (1996) Greek species of
the genus Hydraena (subgenus Haenydra Rey) (Coleoptera,
Hydraenidae). Aquatic Insects, 18, 65–90.
Audisio, P., Trizzino, M., De Biase, A., Mancini, E. & Antonini,
G. (2009) A new species of Hydraena (Coleoptera: Hydra-
enidae) of the H. evanescens complex from Sardinia. Zoo-
taxa, 2318, 281–289.
Barraclough, T.G. & Nee, S. (2001) Phylogenetics and specia-
tion. Trends in Ecology and Evolution, 16, 391–399.
Barraclough, T.G. & Vogler, A.P. (2000) Detecting the geo-
graphical pattern of speciation from species-level phyloge-
nies. The American Naturalist, 155, 419–434.
Bennett, K.D. & Provan, J. (2008) What do we mean by
‘refugia’? Quaternary Science Reviews, 27, 2449–2455.
Berthelemy, C. (1986) Remarks on the genus Hydraena and
revision of the subgenus Phothydraena (Coleoptera:
Hydraenidae). Annales de Limnologie, 22, 181–193.
Beutel, R.G., Anton, E. & Jach, M.A. (2003) On the evolution
of adult head structures and the phylogeny of Hydraenidae
(Coleoptera, Staphyliniformia). Journal of Zoological Sys-
tematics and Evolutionary Research, 41, 256–275.
Bonnet, E. & Van de Peer, Y. (2002) zt: a software tool for
simple and partial Mantel tests. Journal of Statistical Soft-
ware, 7, 101–112.
Butlin, R.K., Galindo, J. & Grahame, J.W. (2008) Sympatric,
parapatric or allopatric: the most important way to classify
speciation? Philosophical Transactions of the Royal Society B:
Biological Sciences, 363, 2997–3007.
Carranza, S. & Wade, E. (2004) Taxonomic revision of Algero-
Tunisian Pleurodeles (Caudata: Salamandridae) using
molecular and morphological data. Revalidation of the
taxon Pleurodeles nebulosus (Guichenot, 1850). Zootaxa,
488, 1–24.
Carranza, S., Romano, A., Arnold, E.N. & Sotgiu, G. (2008)
Biogeography and evolution of European cave salaman-
ders, Hydromantes (Urodela: Plethodontidae), inferred
from mtDNA sequences. Journal of Biogeography, 35, 724–
738.
Chesser, R.T. & Zink, R.M. (1994) Modes of speciation in birds
– a test of Lynch’s method. Evolution, 48, 490–497.
Coyne, J.A. & Orr, H.A. (2004) Speciation. Sinauer Associates,
Sunderland, MA.
Drummond, A. & Rambaut, A. (2007) BEAST: Bayesian evo-
lutionary analysis by sampling trees. BMC Evolutionary
Biology, 7, 214.
Dynesius, M. & Jansson, R. (2000) Evolutionary consequences
of changes in species’ geographical distributions driven by
Milankovitch climate oscillations. Proceedings of the
National Academy of Sciences USA, 97, 9115–9120.
I. Ribera et al.
514 Journal of Biogeography 38, 502–516ª 2010 Blackwell Publishing Ltd
Fitzpatrick, B.M. & Turelli, M. (2006) The geography of
mammalian speciation: mixed signals from phylogenies and
range maps. Evolution, 60, 601–615.
Fresneda, J., Lagar, A., Sainz-Cantero, C.E. & Delgado, J.A.
(1994) Sobre las especies ibericas del grupo de Hydraena
tatii (Coleoptera, Hydraenidae). Zoologica Baetica, 5,
27–39.
Ganglbauer, L. (1901) Beitrage zur Kenntniss der palaarktis-
chen Hydrophiliden. Verhandlungen der Zoologisch-Botan-
ischen Gesellschaft in Wien, 51, 312–332.
Garcıa-Castellanos, D., Estrada, F., Jimenez-Munt, I., Gorini,
C., Fernandez, M., Verges, J. & De Vicente, R. (2009) Cat-
astrophic flood of the Mediterranean after the Messinian
salinity crisis. Nature, 462, 778–781.
Gaston, K.J. (1998) Species-range size distributions: products
of speciation, extinction and transformation. Philosophical
Transactions of the Royal Society B: Biological Sciences, 353,
219–230.
Gaston, K.J. (2003) The structure and dynamics of geographic
ranges. Oxford University Press, Oxford.
Gillespie, R.G. & Roderick, G.K. (2002) Arthropods on islands:
colonization, speciation, and conservation. Annual Review of
Entomology, 47, 595–632.
Gomez, A. & Lunt, D.H. (2007) Refugia within refugia: pat-
terns of phylogeographic concordance in the Iberian Pen-
insula. Phylogeography of southern European refugia (ed. by S.
Weiss and N. Ferrand), pp. 155–188. Springer, Dordrecht.
Graham, R.L. & Hell, P. (1985) On the history of the mini-
mum spanning tree problem. Annals of the History of
Computing, 7, 43–57.
Hansen, M. (1998) World catalogue of insects. Vol. 1. Hydrae-
nidae (Coleoptera). Apollo Books, Stenstrup, Denmark.
Hennig, W. (1966) Phylogenetic systematics. University of Illi-
nois Press, Urbana.
Hewitt, G. (2000) The genetic legacy of the Quaternary ice
ages. Nature, 405, 907–913.
Huelsenbeck, J.P. & Ronquist, F. (2001) MrBayes: Bayesian
inference of phylogenetic trees. Bioinformatics, 17, 754–755.
Ienistea, M.A. (1968) Die Hydraeniden Rumaniens (Coleop-
tera, Hydraenidae). Travaux du Museum National d’historie
naturelle Grigore Antipa, 8, 759–795.
Jablonski, D. (1987) Heritability at the species level: analysis of
geographic ranges of Cretaceous mollusks. Science, 238,
360–363.
Jach, M.A. (1995) The Hydraena (Haenydra) gracilis Germar
species complex (Insecta: Coleoptera: Hydraenidae). Anna-
len des Naturhistorischen Museums in Wien, Series B, 97,
177–190.
Jach, M.A. (2004) Hydraenidae. Catalogue of Palaearctic
Coleoptera. Vol. 2: Hydrophiloidea – Histeroidea – Staphyli-
noidea (ed. by I. Lobl and A. Smetana), pp. 102–122. Apollo
Books, Stenstrup, Denmark.
Jach, M.A. & Balke, M. (2008) Global diversity of water beetles
(Coleoptera) in freshwater. Hydrobiologia, 595, 419–442.
Jach, M.A., Beutel, R.G., Dıaz, J.A. & Kodada, J. (2000) Sub-
generic classification, description of head structures, and
world check list of Hydraena Kugelann (Insecta: Coleoptera:
Hydraenidae). Annalen des Naturhistorischen Museums in
Wien, Series B, 102, 177–258.
Janssens, E.L. (1965) Les Hydraena de l’Egeide. Memoires de
l’Academie Royale de Belgique, Classe des Sciences, 16,
1–126.
Katoh, K. & Toh, H. (2008) Recent developments in the
MAFFT multiple sequence alignment program. Briefings in
Bioinformatics, 9, 286–298.
Keppel, G., Lowe, A.J. & Possingham, H.P. (2009) Changing
perspectives on the biogeography of the tropical south Pa-
cific: influences of dispersal, vicariance and extinction.
Journal of Biogeography, 36, 1035–1054.
Kimura, M. (1953) ‘‘Stepping-stone’’ model of population.
Annual Report of the National Institute of Genetics, Japan, 3,
62–63.
Lambeck, K. & Chappell, J. (2001) Sea level change through the
last glacial cycle. Science, 292, 679–686.
Lisiecki, L.E. & Raymo, M.E. (2007) Plio-Pleistocene climate
evolution: trends and transitions in glacial cycle dynamics.
Quaternary Science Reviews, 26, 56–69.
Losos, J. & Glor, R.E. (2003) Phylogenetic comparative
methods and the geography of speciation. Trends in Ecology
and Evolution, 18, 220–227.
Lynch, J.D. (1989) The gauge of speciation: on the frequencies
of modes of speciation. Speciation and its consequences (ed.
by D. Otte and J.A. Endler), pp. 527–553. Sinauer Associ-
ates, Sunderland, MA.
MacArthur, R.H. & Wilson, E.O. (1967) The theory of island
biogeography. Princeton University Press, Princeton, NJ.
Martınez-Solano, I., Goncalves, H.A., Arntzen, J.W. & Garcıa-
Parıs, M. (2004) Phylogenetic relationships and biogeogra-
phy of midwife toads (Discoglossidae: Alytes). Journal of
Biogeography, 31, 603–618.
Mayr, E. (1963) Animal species and evolution. Harvard Uni-
versity Press, Cambridge, MA.
Mijarra, J.M.P., Barron, E., Manzaneque, F.G. & Morla, C.
(2009) Floristic changes in the Iberian Peninsula and Bale-
aric islands (south-west Europe) during the Cenozoic.
Journal of Biogeography, 36, 2025–2043.
Moritz, C., Patton, J.L., Schneider, C.J. & Smith, T.B. (2000)
Diversification of rainforest faunas: an integrated molecular
approach. Annual Review of Ecology and Systematics, 31,
533–563.
d’Orchymont, A. (1925) Contribution a l’etude des hydro-
philides. IV. Bulletin et Annales de la Societe Royale d’Ento-
mologie de Belgique, 65, 298–305.
d’Orchymont, A. (1935) Inventa entomologica itineris hispa-
nici et maroccani quod a. 1926 fecerunt Harald ot Hakan
Lindberg. XXII. Palpicornia. Commentationes Biologicae,
Helsingfors, 5, 1–22.
Perkins, P.D. (1989) Adelphydraena, new genus, and two new
species from Venezuela, and remarks on phylogenetic
relationships within the subtribe Hydraenina (Coleoptera:
Hydraenidae). Proceedings of the Biological Society of
Washington, 102, 447–457.
The geography of speciation in the Haenydra lineage
Journal of Biogeography 38, 502–516 515ª 2010 Blackwell Publishing Ltd
Perkins, P.D. (1997) Life on the effective bubble: exocrine
secretion delivery systems (ESDS) and the evolution and
classification of beetles in the family Hydraenidae (Insecta:
Coleoptera). Annals of the Carnegie Museum, 66, 89–207.
Perkins, P.D. (in press) New species (130) of the megadiverse
aquatic beetle genus Hydraena Kugelann from Papua New
Guinea, and a preliminary analysis of areas of endemism
(Coleoptera: Hydraenidae). Zootaxa.
Rey, C. (1886) Histoire naturelle des Coleopteres de France
(suite). Annales de la Societe linneenne de Lyon, 32(1885),
1–187.
Ribera, I. & Vogler, A.P. (2004) Speciation of Iberian diving
beetles in Pleistocene refugia (Coleoptera, Dytiscidae).
Molecular Ecology, 13, 179–193.
Ribera, I., Fresneda, J., Bucur, R., Izquierdo, A., Vogler, A.P.,
Salgado, J.M. & Cieslak, A. (2010a) Ancient origin of a
western Mediterranean radiation of subterranean beetles.
BMC Evolutionary Biology, 10, 29.
Ribera, I., Castro, A. & Hernando, C. (2010b) Ochthebius
(Enicocerus) aguilerai sp.n. from central Spain, with a
molecular phylogeny of the Western Palaearctic species of
Enicocerus (Coleoptera, Hydraenidae). Zootaxa, 2351, 1–13.
Rocchi, S. (2009) Coleotteri acquatici endemici presenti
nell’appennino settentrionale (Coleoptera: Dytiscidae,
Hydraenidae, Scirtidae, Elmidae). Onychium, 7, 48–52.
Sanchez-Fernandez, D., Bilton, D.T., Abellan, P., Ribera, I.,
Velasco, J. & Millan, A. (2008) Are the endemic water beetles
of the Iberian Peninsula and the Balearic islands effectively
protected? Biological Conservation, 141, 1612–1627.
Sanmartın, I. (2003) Dispersal vs. vicariance in the Mediter-
ranean: historical biogeography of the Palearctic Pachy-
deminae (Coleoptera, Scarabaeoidea). Journal of
Biogeography, 30, 1883–1897.
Stamatakis, A., Hoover, P. & Rougemont, J. (2008) A rapid
bootstrap algorithm for the RAxML web servers. Systematic
Biology, 57, 758–771.
Stewart, J.R., Lister, A.M., Barnes, I. & Dalen, L. (2010) Refugia
revisited: individualistic responses in space and time.
Proceedings of the Royal Society B: Biological Sciences, 277,
661–671.
Suc, J.P. (1984) Origin and evolution of the Mediterranean
vegetation and climate in Europe. Nature, 307, 429–432.
Swofford, D.L. (2002) PAUP*: phylogenetic analysis using
parsimony (*and other methods), Version 4. Sinauer,
Sunderland, MA.
Wiens, J.J. (2004) Speciation and ecology revisited: phyloge-
netic niche conservatism and the origin of species. Evolution,
58, 193–197.
Wright, S. (1943) Isolation by distance. Genetics, 28, 114–
138.
Zangari, F., Cimmaruta, R. & Nascetti, G. (2006) Genetic rela-
tionships of the western Mediterranean painted frogs based
on allozymes and mitochondrial markers: evolutionary and
taxonomic inferences (Amphibia, Anura, Discoglossidae).
Biological Journal of the Linnean Society, 87, 515–536.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the
online version of this article:
Appendix S1 Specimens used in the study, with locality,
collector, voucher reference numbers and accession numbers
for the sequences.
Appendix S2 List of primers used for sequencing.
As a service to our authors and readers, this journal
provides supporting information supplied by the authors.
Such materials are peer-reviewed and may be re-organized
for online delivery, but are not copy-edited or typeset.
Technical support issues arising from supporting informa-
tion (other than missing files) should be addressed to the
authors.
BIOSKETCH
I. Ribera is interested in the evolution, systematics and
biogeography of Coleoptera. This work is part of an ongoing
collaboration among the authors to study the Iberian species
of Hydraena.
Author contributions: I.R. conceived the study; I.R., A.C., J.G.,
L.F.V. and M.A.J. provided material and data; A.I. did most of
the laboratory work; I.R. carried out the phylogenetic analyses
and outlined a first draft; A.C., J.G., J.A.D., L.F.V. and M.A.J.
discussed the results and contributed to the final version of the
manuscript.
Editor: Pauline Ladiges
I. Ribera et al.
516 Journal of Biogeography 38, 502–516ª 2010 Blackwell Publishing Ltd